Research Engineer - Data Eval

Calder•Redwood City, CA

About The Position

Calder is hiring a Research Engineer to own the measurement loop for the quality of real-world and egocentric capture data used to train world models and robotic foundation models. The role evaluates data from raw capture through downstream model impact, including measuring how collection, curation, and labeling choices affect performance and comparison with public baselines. The position involves training, fine-tuning, and benchmarking open-source world models and robotic policies, as well as validating raw data to detect low-quality, corrupted, or out-of-distribution captures. It also helps shape Calder's evaluation research agenda, develops benchmarks and metrics, runs end-to-end model evaluation, and builds automated pipelines and tooling. The engineer investigates corner cases, failure modes, and safety, privacy, and risk issues, sharing findings with research and robotics teams. The role includes technical leadership and mentorship as Calder builds its evaluation team.

Requirements

  • PhD in Computer Science, Machine Learning, Robotics, Computer Vision, or a closely related discipline.
  • A record of shipping high-impact research in fast-moving settings, supported by first-author papers at venues such as NeurIPS, ICML, ICLR, CVPR, CoRL, RSS, ICRA, or IROS.
  • Strong mathematical foundations and depth in at least one area relevant to physical AI evaluation, such as reward modeling for embodied agents, calibration and uncertainty quantification, simulation-to-real validity, or human-centered AI methods for spatial data.
  • Hands-on experience with modern ML systems in research and production, including meaningful exposure to evaluating physical AI such as robotic manipulation, navigation, world models, vision-language-action models, or multimodal egocentric perception.
  • Fluency in Python and strong engineering skills, including building scalable data pipelines and evaluation harnesses for simulation and real-world deployments, and defining metrics for embodied model quality.
  • Comfort with the ML experimentation stack, including PyTorch, training and evaluation infrastructure, and open-source model ecosystems.
  • Rigor in experimental design.

Nice To Haves

  • Experience with robotics, imitation learning, world models, computer vision models, or large-scale data pipelines.
  • A PhD thesis or research focus related to evaluation, such as reward modeling, preference learning, human feedback, calibration, or automated assessment for embodied or perception systems.

Responsibilities

  • Train, fine-tune, and benchmark open-source world models and robotic policies to measure how collection, curation, and labeling choices affect downstream performance and compare results with public baselines.
  • Validate raw collected data using models to flag low-quality, corrupted, or out-of-distribution captures, including defects such as motion blur, exposure problems, dropped frames, sensor desynchronization, calibration drift, and SLAM or stereo-depth failures.
  • Help define Calder's evaluation research agenda and run disciplined experimental programs addressing important problems in embodied and physical AI.
  • Develop benchmarks, metrics, and methodologies for evaluating how robotic and world models sense, reason, and act.
  • Run model evaluation end to end, including offline studies, A/B and regression testing, and in-the-wild deployment across sensor stacks, long-tail edge cases, and real-world manipulation, navigation, and interaction tasks.
  • Build pipelines, dashboards, and tools to make evaluation fast, reproducible, and largely automated for internal teams and external partners.
  • Identify corner cases, failure modes, and safety, privacy, and risk issues, and communicate findings to research and robotics teams.
  • Provide technical leadership and mentorship across cross-functional efforts while helping build the evaluation team.
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service